CN112247673B - Woodworker cutter wear state diagnosis method based on genetic BP neural network - Google Patents

Woodworker cutter wear state diagnosis method based on genetic BP neural network Download PDF

Info

Publication number
CN112247673B
CN112247673B CN202010926651.9A CN202010926651A CN112247673B CN 112247673 B CN112247673 B CN 112247673B CN 202010926651 A CN202010926651 A CN 202010926651A CN 112247673 B CN112247673 B CN 112247673B
Authority
CN
China
Prior art keywords
wear state
tool
neural network
cutter
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010926651.9A
Other languages
Chinese (zh)
Other versions
CN112247673A (en
Inventor
胡勇
田广军
郭晓磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Bosheng Prewi Shanghai Tools Co ltd
Nanjing Forestry University
Original Assignee
Bosheng Prewi Shanghai Tools Co ltd
Nanjing Forestry University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Bosheng Prewi Shanghai Tools Co ltd, Nanjing Forestry University filed Critical Bosheng Prewi Shanghai Tools Co ltd
Priority to CN202010926651.9A priority Critical patent/CN112247673B/en
Publication of CN112247673A publication Critical patent/CN112247673A/en
Application granted granted Critical
Publication of CN112247673B publication Critical patent/CN112247673B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
    • B23Q17/00—Arrangements for observing, indicating or measuring on machine tools
    • B23Q17/09—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool
    • B23Q17/0952—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool during machining
    • B23Q17/0957—Detection of tool breakage
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B27—WORKING OR PRESERVING WOOD OR SIMILAR MATERIAL; NAILING OR STAPLING MACHINES IN GENERAL
    • B27C—PLANING, DRILLING, MILLING, TURNING OR UNIVERSAL MACHINES FOR WOOD OR SIMILAR MATERIAL
    • B27C5/00—Machines designed for producing special profiles or shaped work, e.g. by rotary cutters; Equipment therefor
    • B27C5/02—Machines with table
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Mechanical Engineering (AREA)
  • Physiology (AREA)
  • Genetics & Genomics (AREA)
  • Wood Science & Technology (AREA)
  • Forests & Forestry (AREA)
  • Numerical Control (AREA)

Abstract

The invention discloses a woodworking tool wear state diagnosis method based on a genetic BP neural network, which comprises the following steps: the method comprises the following steps: the acquisition comprises the front angle gamma of a cutter, the rotating speed n of a main shaft, the cutting depth h and the power P of a main shaft of a machine tooltA sample set including the approximation coefficient of (a) and tool wear (VB); step two: and establishing a tool wear state diagnosis model by using the genetic BP neural network. The technology of the invention uses the power signal as the tool wear state diagnosis signal, reduces the sensor cost, solves the sensor installation problem, ensures that the signal precision is not interfered by the processing environment, simultaneously considers the front angle gamma of the tool, the rotating speed n of the main shaft and the cutting depth h, ensures that the model can be suitable for various cutting parameters, establishes the tool wear state diagnosis model by using the genetic BP neural network of the relu training function, and greatly improves the calculation speed and the diagnosis precision of the BP neural network.

Description

Woodworker cutter wear state diagnosis method based on genetic BP neural network
Technical Field
The invention relates to the field of cutter wear state diagnosis, in particular to a woodworking cutter wear state diagnosis method based on a genetic BP neural network.
Background
The woodworking tool is an important component in the manufacturing process of the wood product, and is influenced by factors such as processing parameters, machine tool performance and the like in the cutting process to cause tool abrasion, so that the edge surface of the wood product is torn, cut and the like, and the surface roughness and the integral attractiveness of the wood product are influenced. The adoption of accurate and reliable cutter wear state diagnosis technology can improve the utilization rate of the numerical control machine by 50 percent, reduce the production cost by about 30 percent, improve the product quality and reduce the defective rate of products.
At present, the tool wear state diagnosis technology is mainly divided into two types, one type is that the relation between various signal characteristic values and tool wear state is analyzed by collecting characteristic values of various signals related to the tool wear state, a tool wear threshold value is established, the tool wear state is diagnosed, for example, Chen Guassian et al collects power signals of a machine tool spindle, extracts a signal entropy value, establishes a tool wear threshold value, and diagnoses the tool wear state (authorization number: CN108490880B), but the technology ignores the problem that the tool wear threshold value can be changed when cutting parameters are changed; the other method is that after signals related to the wear state of the cutter are collected, a cutter wear state diagnosis model is established to realize cutter wear state diagnosis, such as image features extracted by Guanshan and the like and fused with collected acoustic emission signals, a cutter wear state classifier is established by using a discrete hidden Markov model, and the cutter wear state is automatically diagnosed (authorization number: CN 107378641B).
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a woodworking tool wear state diagnosis method based on a BP neural network, so as to solve the problems mentioned in the background art.
The invention provides a woodworking tool wear state diagnosis method based on a genetic BP neural network, which comprises the following steps:
the method comprises the following steps: collecting a sample set including a tool front angle gamma, a spindle rotating speed n, a cutting depth h, an approximate coefficient of machine tool spindle power Pt and tool abrasion VB;
step two: establishing a tool wear state diagnosis model by using a genetic BP neural network;
step three: establishing a tool wear state diagnosis standard;
step four: transmitting real-time data to enter a tool wear state diagnosis model and diagnosing the tool wear state by combining with a tool wear state diagnosis standard;
the method for establishing the tool wear state diagnosis model in the second step comprises the following steps:
21): dividing a sample set into a training set and a testing set;
22): carrying out normalization processing on the training set and the test set;
23): constructing a BP neural network;
24): solving the optimal initial threshold value and weight of the BP neural network by using a genetic algorithm;
25): training a BP neural network to obtain a diagnosis model;
the real-time data in the fourth step comprise a tool front angle gamma, a main shaft rotating speed n, a cutting depth h and a machine tool main shaft power PtThe approximation coefficient of (c).
Preferably, in the step one, a tool angle γ, a spindle speed n and a cutting depth h are selected as tool wear state diagnosis sample set parameters, and since cutting parameters affecting tool wear VB in wood cutting include a feed speed f, a tool angle γ, a spindle speed n and a cutting depth h, the feed speed f is usually kept constant, and tool wear state diagnosis in different cutting parameters can be realized by only changing the tool angle γ, the spindle speed n and the cutting depth h to acquire tool wear state diagnosis signals.
Preferably, in the step one, the machine tool spindle power Pt is selected as a tool wear state diagnosis sample set parameter, and as the machine tool spindle power Pt is directly linked with the cutting force F, after the tool is worn, the contact area and friction between the tool and a workpiece are increased, so that the cutting force F is increased, and the current and load power of a machine tool spindle motor are increased, the machine tool spindle motor power Pt can accurately reflect the tool wear state, so that the tool wear state diagnosis precision is further improved, and the power sensor is relatively cheap and easy to install, and only needs to be accessed into a machine tool control box, and is not influenced by a machine tool workpiece clamping mode.
Preferably, in the step one, the power sensor is firstly connected to the output end of the machine tool control box, and then the machine tool spindle power Pt is collected according to a full-factor experimental scheme, wherein the full-factor experimental scheme comprises different cutting parameters and tool wear VB.
Preferably, in step one, in order to extract the characteristics related to the wear state of the tool in the machine tool spindle power Pt and remove noise and interference in the signal, the machine tool spindle power Pt is subjected to characteristic extraction by using Discrete Wavelet Transform (DWT) with a wavelet function of DB4, and the extracted characteristics are approximate coefficients of the discrete wavelet transform.
Preferably, the feature extraction of the step one and the model establishment of the step two are performed on software MATLAB.
Preferably, in step 21), approximate coefficients of a tool rake angle γ, a spindle rotation speed n, a cutting depth h, and a machine tool spindle power Pt are used as input vectors in the BP neural network, and tool wear VB is used as an output vector.
Preferably, in step 22), in order to make the BP neural network converge quickly and avoid repeated values, the normalization process needs to perform normalization on the sample set, the normalization process selects linear normalization, and normalizes all input and output data into a range of [0, 1], where the processing formula is as follows:
yi＝(xi-xmin)÷(xmax-xmin)
in the formula, yiFor the sample set after processing, xiFor the original sample set, xmaxIs the maximum value of the original sample set, xminIs the original sample set minimum.
Preferably, in step 24), a genetic algorithm is used to perform global optimization search on the initial threshold and the weight of the BP neural network, so as to avoid the difficult problem that the BP neural network is trapped in local optimization, improve the diagnosis precision of the BP neural network, and the genetic algorithm needs to set initial parameters, including: the population scale, fitness function, evolution times, cross probability and variation probability, wherein the population scale selection range is 30-50, the evolution times selection range is 30-100, the cross probability selection range is 0.01-1, and the variation probability selection range is 0.01-1, in order to distinguish the advantages and disadvantages of individuals in a genetic algorithm, the fitness is used for measuring the survival degree of the individuals in the population, the fitness is calculated through the fitness function, when the genetic algorithm tends to converge, as the individual fitness difference in the population is small, the potential of continuous optimization is reduced, the local optimal solution is possibly trapped, and the problem can be solved by selecting the appropriate fitness function, and the fitness function is selected as follows:
Figure BDA0002668589140000041
in the formula, YiFor data predicted by BP neural networks, XiFor real data, N represents the number of sample sets, and ε is a constant that avoids a denominator of 0.
Preferably, the training of the tool wear state diagnostic model in step 25) requires setting the following parameters, including the number of hidden layers, training times, learning rate, training precision and training function, wherein the number of hidden layers, training times, learning rate and training precision are generally selected according to experience, the range of the training times is 50-200, the learning rate is 0.01, the training precision is 0.0001, and the number of hidden layers is selected according to the following formula:
Figure BDA0002668589140000042
in the formula, m is the number of hidden layers, n is the number of input vectors, j is the number of output vectors, alpha is a constant between 1 and 10, the value of the specific hidden layers needs to be verified one by one in the value range of m, the highest prediction result is taken as the actual hidden layer, the relu function is selected by the training function, and compared with the common sigmoid function, the relu function can prevent network gradient dispersion and accelerate the calculation speed of the BP neural network, so that the accuracy of the BP neural network is improved, and the formula is as follows:
f(x)＝max(0,x)
in the formula, x represents training set data, and when x is smaller than 0, the output value is 0, and when x is larger than 0, the output value is x.
Preferably, the tool wear state diagnostic criteria in step three are as follows: for the wear VB of the output cutter, when the wear VB of the rear cutter face is less than or equal to 0.1mm, the cutter is in a brand-new state, when the wear VB is more than 0.1mm and less than or equal to 0.2mm, the cutter is in a normal wear state, when the wear VB is more than 0.2mm and less than or equal to 0.3mm, the cutter is in a rapid wear state, and when the wear VB is more than 0.3mm, the cutter is in a damaged state.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a woodwork cutter wear state diagnosis method based on a genetic BP neural network, which comprises the steps of firstly collecting the power P of a main shaft of a machine tool according to a front angle gamma of a cutter, the rotating speed n of the main shaft and the cutting depth htThe diagnostic model is used as a tool wear state diagnostic related signal, a BP neural network is optimized through a genetic algorithm to train and optimize to establish a tool wear state diagnostic model, and meanwhile, a tool wear state judgment criterion is established, so that the diagnosis of the woodworking tool wear state through cutting parameters and cutting power is realized.
The technology of the invention uses the power signal as the tool wear state diagnosis signal, reduces the cost of the sensor, solves the problem of sensor installation, ensures that the signal precision is not interfered by the processing environment, simultaneously considers the front angle gamma of the tool, the rotating speed n of the main shaft and the cutting depth h, ensures that the model can be suitable for various cutting parameters, establishes the tool wear state diagnosis model by using the genetic BP neural network of the relu training function, greatly improves the calculation speed and the diagnosis precision of the BP neural network, effectively avoids the gradient diffusion of the BP neural network, and further improves the practicability of the technology of the invention.
Drawings
FIG. 1 is a flow chart of a method for diagnosing the wear state of a woodworking tool based on a genetic BP neural network in the invention;
FIG. 2 is a schematic diagram of a machine tool spindle power acquisition device according to the present invention;
FIG. 3 is a sample set acquisition method of the present invention;
FIG. 4 is a flow chart of the steps of the present invention for building a diagnostic model using a genetic BP neural network;
FIG. 5 illustrates a method for identifying the wear state of a woodworking tool;
FIG. 6 is an error analysis of the present invention.
The figures in the drawings are marked with numbers: 1-machine tool spindle motor; 2-milling cutter; 3-a workpiece; 4-vacuum adsorption stage; 5-a machine tool control box; 6-power analyzer; 7-data processing computer.
Detailed Description
In order that the objects and advantages of the invention will be more clearly understood, the invention is further described in detail below with reference to examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In order to prove the accuracy of the diagnosis method of the embodiment, a verification test was performed in a numerical control machining center of a certain wood factory. Milling a wood-plastic composite (WPC) by using a shank milling cutter with the diameter of 18mm, and acquiring machine tool spindle power P by using a power analyzer WT500 according to experimental cutting parameters in a table 2tAnd establishing a sample set.
As shown in fig. 1, the method for diagnosing the wear state of a woodworking tool based on the machine tool spindle power and the genetic BP neural network of the present invention comprises the following steps:
the method comprises the following steps: a sample set is collected.
As shown in figure 2, a workpiece (3) to be cut is adsorbed on a machine tool vacuum adsorption table (4), a machine tool spindle motor (1) drives a milling cutter (2) to cut the workpiece (3), and meanwhile, a power analyzer (6) connected to the output end of the spindle motor of a machine tool control box (5) is used for collecting the power P of the machine tool spindle of the cuttingtAnd transmits the collected data to a data processing computer (7) for data processing.
As shown in FIG. 3, the tool rake angle γ, the spindle speed n, the cutting depth h, and the machine tool spindle power P are measured for each cuttingtThe approximate coefficient and the tool wear VB are used as sample data, and cutting is repeated according to the cutting parameters in the table 1 and the full-factor scheme until all sample sets are collected.
TABLE 1 test cutting parameters
Figure BDA0002668589140000061
In order to extract the characteristics related to the wear state of the tool in the machine tool spindle power Pt and remove noise and interference in signals, the machine tool spindle power Pt is subjected to characteristic extraction using Discrete Wavelet Transform (DWT) with a wavelet function of DB4, the discrete wavelet transform decomposes the machine tool spindle power Pt 5 times, the extracted characteristics are approximation coefficients of the discrete wavelet transform, and table 2 is a partially extracted characteristic approximation coefficient.
TABLE 2 characteristic approximation coefficients
Figure BDA0002668589140000071
Step two: and establishing a tool wear state diagnosis model by using the genetic BP neural network.
As shown in fig. 4, a tool wear state diagnostic model is established on MATLAB, and the specific method is as follows:
21): dividing a sample set into a training set and a testing set;
22): carrying out normalization processing on the training set and the test set;
23): constructing a BP neural network;
24): solving the optimal initial threshold value and weight of the BP neural network by using a genetic algorithm;
25): and training the BP neural network to obtain a diagnosis model.
The specific algorithm is shown in table 3:
TABLE 3 tool wear State diagnostic model implementation Algorithm
Figure BDA0002668589140000072
Figure BDA0002668589140000081
Figure BDA0002668589140000091
Figure BDA0002668589140000101
Figure BDA0002668589140000111
Wherein, approximate coefficients of a tool front angle gamma, a spindle rotating speed n, a cutting depth h and a machine tool spindle power Pt in the step 21) are used as input vectors in a BP neural network, and tool wear VB is used as an output vector;
in step 21), xtrin represents 46 training sets, and Xtest represents 8 testing sets, and the total number of the testing sets is 54 sample sets.
In step 22), in order to make the BP neural network converge quickly and avoid repeated values, normalization processing needs to be performed on the sample set, so that the S-type activation function is used to perform normalization processing on the input and output data, and all the input and output data are normalized to the range of [0, 1 ]. The activation function of the S-type is formulated as follows:
yi＝(xi-xmin)÷(xmax-xmin)
in the formula, yiFor the sample set after processing, xiFor the original sample set, xmaxIs the maximum value of the original sample set, xminIs the original sample set minimum.
And 24) performing global optimization search on the initial threshold and the weight of the BP neural network by using a genetic algorithm, avoiding the difficult problem that the BP neural network is trapped in local optimum, improving the diagnosis precision of the BP neural network, wherein the genetic algorithm needs to set initial parameters and comprises the following steps: the method comprises the following steps of population scale, fitness function, evolution times, cross probability and variation probability, wherein the population scale is selected to be 40, the evolution times are selected to be 40, the cross probability is selected to be 0.8, the variation probability is selected to be 0.2, and the fitness function is as follows:
Figure BDA0002668589140000121
in the formula, YiFor data predicted by BP neural networks, XiFor real data, N represents the number of sample sets, and ε is a constant that avoids the denominator being 0;
in the step 25), the following parameters including the number of hidden layers, the number of training times, the learning rate, the training precision and the training function are required to be set for training the tool wear state diagnosis model, wherein the number of training times is selected to be 100, the learning rate is selected to be 0.01, the training precision is selected to be 0.0001, and the number of hidden layers is selected according to the following formula:
Figure BDA0002668589140000122
in the formula, m is the number of hidden layers, n is the number of input vectors, n is 7, j is the number of output vectors, j is 1, alpha is a constant between 1 and 10, the value range of m is 4 to 13, the value of the specific hidden layers is verified one by one in the value range of m, the highest prediction result is taken as the actual hidden layer number, m is 10 after verification, a relu function is selected by a training function, and the relu function can prevent network gradient diffusion and accelerate the calculation speed of a BP neural network compared with a common sigmoid function, so that the accuracy of the BP neural network is improved, and the formula is as follows:
f(x)＝max(0,x)
in the formula, x represents sample training set data, when x is smaller than 0, the output value is 0, and when x is larger than 0, the output value is x;
step three: and establishing a tool wear state diagnosis standard.
For the wear VB of the output cutter, when the wear VB of the rear cutter face is less than or equal to 0.1mm, the cutter is in a brand-new state, when the wear VB is more than 0.1mm and less than or equal to 0.2mm, the cutter is in a normal wear state, when the wear VB is more than 0.2mm and less than or equal to 0.3mm, the cutter is in a rapid wear state, and when the wear VB is more than 0.3mm, the cutter is in a damaged state.
Step four: and transmitting real-time data to enter a tool wear state diagnosis model and diagnosing the tool wear state by combining with a tool wear state diagnosis standard.
As shown in fig. 5, the method for identifying the wear state of a woodworking tool according to the present invention includes inputting real-time data including a tool rake angle γ, a spindle rotation speed n, a cutting depth h, and an approximation coefficient of a machine tool spindle power Pt into a tool wear state diagnosis model, comparing a tool wear VB obtained from the output model diagnosis result with a tool wear state diagnosis standard, and obtaining a tool wear state according to the comparison result.
The acquired sample set is used for training and optimizing to establish a tool wear diagnosis model, then 8 groups of data are randomly acquired to verify the model accuracy, as shown in fig. 6, model error analysis shows that the model prediction condition basically accords with the actual condition, and the calculation accuracy is 97%, so that the tool wear state diagnosis method is high in accuracy and can be used for practical production application.

Claims (5)

1. A woodworking tool wear state diagnosis method based on a genetic BP neural network is characterized by comprising the following steps:
the method comprises the following steps: the acquisition comprises the front angle gamma of a cutter, the rotating speed n of a main shaft, the cutting depth h and the power P of a main shaft of a machine tooltA sample set including the approximation coefficient of (a) and tool wear (VB);
step two: establishing a tool wear state diagnosis model by using a genetic BP neural network;
step three: establishing a tool wear state diagnosis standard;
step four: transmitting real-time data to enter a tool wear state diagnosis model and diagnosing the tool wear state by combining with a tool wear state diagnosis standard;
and in the second step, the establishment method of the tool wear state diagnosis model comprises the following steps:
21): dividing a sample set into a training set and a testing set;
22): carrying out normalization processing on the training set and the test set;
23): constructing a BP neural network;
24): solving the optimal initial threshold and weight of the BP neural network by using a genetic algorithm;
25): training a BP neural network to obtain a diagnosis model;
the real-time data in the fourth step comprise a tool front angle gamma, a main shaft rotating speed n, a cutting depth h and a machine tool main shaft power PtThe approximation coefficient of (2);
in the first step, selecting a cutter angle gamma, a main shaft rotating speed n and a cutting depth h as parameters of a cutter wear state diagnosis sample set;
in the first step, a power sensor is firstly connected to the output end of a machine tool control box, and then the power Pt of a main shaft of the machine tool is collected according to a full-factor experimental scheme, wherein the full-factor experimental scheme comprises different cutting parameters and tool wear VB;
in the first step, in order to extract the characteristics related to the wear state of the tool in the machine tool spindle power Pt and remove noise and interference in signals, the acquired machine tool spindle power Pt is subjected to characteristic extraction by using Discrete Wavelet Transform (DWT) with a wavelet function of DB4, the discrete wavelet transform decomposes the machine tool spindle power Pt for 5 times, and the extracted characteristics are approximate coefficients of the discrete wavelet transform;
in the step 21), approximate coefficients of a front angle gamma of the cutter, a rotating speed n of a main shaft, a cutting depth h and power Pt of a main shaft of the machine tool are used as input vectors in a BP neural network, and cutter abrasion VB is used as an output vector;
in step 22), the data of the sample set is normalized, the normalization processing selects linear normalization, all input and output data are normalized to be in a range of [0, 1], and a processing formula is as follows:
yi＝(xi-xmin)÷(xmax-xmin)
in the formula, yiFor the sample set x after processingiFor the original sample set xmaxFor large values of the original sample set, xminIs a small value of the original sample set.
2. The method for diagnosing the wear state of the woodworking tool based on the genetic BP neural network as claimed in claim 1, wherein the feature extraction in the step one and the model building in the step two are performed on software MATLAB.
3. The method for diagnosing the wear state of the woodworking tool based on the genetic BP neural network as claimed in claim 1, wherein the genetic algorithm is used in step 24) to perform global optimization search on the initial threshold and the weight of the BP neural network, and the genetic algorithm needs to set initial parameters, and comprises: the population scale, fitness function, evolution times, cross probability and variation probability, wherein the selection range of the population scale is 30-50, the selection range of the evolution times is 30-100, the selection range of the cross probability is 0.01-1, and the selection range of the variation probability is 0.01-1; the fitness function is chosen as follows:
Figure FDA0003336942380000021
in the formula, YiFor data predicted by BP neural networks, XiFor real data, N represents the number of sample sets, and ε is a constant that avoids a denominator of 0.
4. The method for diagnosing the wear state of the woodworking tool based on the genetic BP neural network as claimed in claim 1, wherein the following parameters including the number of implicit layers, the number of training times, the learning rate, the training precision and the training function are required to be set for training the tool wear state diagnostic model in the step 25), wherein the number of implicit layers, the number of training times, the learning rate and the training precision are selected according to experience, the range of the number of training times is 50-200, the range of the learning rate is 0.01, the range of the training precision is 0.0001, and the number of implicit layers is selected according to the following formula:
Figure FDA0003336942380000022
in the formula, m is the number of hidden layers, n is the number of input vectors, j is the number of output vectors, alpha is a constant between 1 and 10, the value of the specific hidden layers needs to be verified one by one in the value range of m, the highest prediction result is taken as the actual hidden layer, the relu function is selected by the training function, the network gradient dispersion can be prevented by the relu function compared with the common sigmoid function, the calculation speed of the BP neural network is accelerated, and the accuracy of the network is improved, wherein the formula is as follows:
f(x)＝max(0,x)
in the formula, x represents training set data, and when x is smaller than 0, the output value is 0, and when x is larger than 0, the output value is x.
5. The method for diagnosing the wear state of the woodworking tool based on the genetic BP neural network as claimed in claim 1, wherein the tool wear state diagnostic criteria in the third step are as follows: for the wear VB of the output cutter, when the wear VB of the rear cutter face is less than or equal to 0.1mm, the cutter is in a brand-new state, when the wear VB is more than 0.1mm and less than or equal to 0.2mm, the cutter is in a normal wear state, when the wear VB is more than 0.2mm and less than or equal to 0.3mm, the cutter is in a rapid wear state, and when the wear VB is more than 0.3mm, the cutter is in a damaged state.
CN202010926651.9A 2020-09-07 2020-09-07 Woodworker cutter wear state diagnosis method based on genetic BP neural network Active CN112247673B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010926651.9A CN112247673B (en) 2020-09-07 2020-09-07 Woodworker cutter wear state diagnosis method based on genetic BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010926651.9A CN112247673B (en) 2020-09-07 2020-09-07 Woodworker cutter wear state diagnosis method based on genetic BP neural network

Publications (2)

Publication Number Publication Date
CN112247673A CN112247673A (en) 2021-01-22
CN112247673B true CN112247673B (en) 2021-12-21

Family

ID=74231804

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010926651.9A Active CN112247673B (en) 2020-09-07 2020-09-07 Woodworker cutter wear state diagnosis method based on genetic BP neural network

Country Status (1)

Country Link
CN (1) CN112247673B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113408182B (en) * 2021-07-16 2022-05-06 山东大学 Tool life full-cycle wear diagnosis method, device and storage medium based on multiple wavelet optimal features and neural network
CN113847949A (en) * 2021-09-23 2021-12-28 徐州万达回转支承有限公司 Multi-working-condition abrasive belt wear state online detection method based on sensor information fusion
CN114200889B (en) * 2021-11-23 2023-10-24 华中科技大学 Milling parameter optimization method based on deep reinforcement learning

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103465107B (en) * 2013-09-24 2016-05-25 沈阳利笙电子科技有限公司 A kind of Tool Wear Monitoring method
CN108107838A (en) * 2017-12-27 2018-06-01 山东大学 A kind of numerical control equipment tool wear monitoring method based on cloud knowledge base and machine learning
CN108389001B (en) * 2018-03-12 2021-10-15 东北大学 Intelligent nonlinear process planning method based on STEP-NC
CN109409568A (en) * 2018-09-19 2019-03-01 安徽农业大学 A kind of prediction technique based on genetic algorithm optimization BP neural network underground water buried depth
CN109146209A (en) * 2018-11-02 2019-01-04 清华大学 Machine tool spindle thermal error prediction technique based on wavelet neural networks of genetic algorithm
CN109434564A (en) * 2018-12-21 2019-03-08 哈尔滨理工大学 A kind of cutter wear state monitoring method based on deep neural network
CN111291518B (en) * 2020-01-21 2024-04-09 上海电气集团股份有限公司 BP neural network-based breaker cutter wear degree judging method

Also Published As

Publication number Publication date
CN112247673A (en) 2021-01-22

Similar Documents

Publication Publication Date Title
CN112247673B (en) Woodworker cutter wear state diagnosis method based on genetic BP neural network
CN111037365B (en) Cutter state monitoring data set enhancing method based on generative countermeasure network
CN111716150B (en) Evolution learning method for intelligently monitoring cutter state
CN109396953B (en) Machine tool working state intelligent identification system based on signal fusion
CN112257530B (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN112247674B (en) Cutter wear prediction method
CN111687689A (en) Cutter wear state prediction method and device based on LSTM and CNN
CN110561191B (en) Numerical control machine tool cutter abrasion data processing method based on PCA and self-encoder
CN102091972A (en) Numerical control machine tool wear monitoring method
CN113927371A (en) Cutter wear prediction method based on multi-sensor feature fusion
CN114619292B (en) Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network
CN110561192A (en) Deep hole boring cutter state monitoring method based on stacking self-encoder
CN111633467A (en) Cutter wear state monitoring method based on one-dimensional depth convolution automatic encoder
CN109605128B (en) Milling chatter online detection method based on power spectrum entropy difference
CN111126255A (en) Numerical control machine tool cutter wear value prediction method based on deep learning regression algorithm
CN113798920A (en) Cutter wear state monitoring method based on variational automatic encoder and extreme learning machine
CN110909710A (en) Self-adaptive main shaft performance degradation identification method based on S test piece
CN113780208A (en) Cutter wear state prediction method and device based on anti-migration learning
CN113627544A (en) Machine tool milling cutter state identification method based on multi-source heterogeneous data fusion
CN113458873A (en) Method for predicting wear loss and residual life of cutter
CN114227382B (en) Cutter damage monitoring system and method based on novel capsule network
CN115587290A (en) Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network
CN115905976A (en) Method, system and equipment for diagnosing high way Bi-LSTM bearing fault based on attention mechanism
CN115221973A (en) Aviation bearing fault diagnosis method based on enhanced weighted heterogeneous ensemble learning
CN114487129A (en) Flexible material damage identification method based on acoustic emission technology

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant